SAN FRANCISCO -- In a change of strategy, Intel Corp. is quietly bypassing an existing design partner to develop its own, next-generation prototype chip, based on a controversial nonvolatile memory technology called ovonic unified memory (OUM).
OUM technology was originally announced in the late-1960s--around the time Intel was formed--but it has not been commercialized due to some technical problems.
Last year, Intel recruited nonvolatile memory design specialist Azalea Microelectronics Corp. to build a prototype 4-megabit chip, based on OUM technology. Santa Clara, Calif.-based Azalea developed a 0.18-micron test chip using OUM, which is one of the potential successors to nonvolatile flash memory (see Nov. 13 story).
Azalea fulfilled its contract to deliver the test chip, but Intel will now take matters into its own hands and develop the next-generation, OUM prototype chip, said Stefan Lai, vice president and co-director of Intel's California Technology and Manufacturing Center. "We will develop the next chip by ourselves," Lai told SBN during an interview at the Intel Developer Forum (IDF) in San Francisco this week.
Intel is working on a next-generation, OUM test chip, which will be based on 0.13-micron process technology. But the company's OUM efforts are still in the R&D stage. There are no plans to commercialize OUM memories until the company can solve some of the ongoing technical problems with the technology, Lai said.
"OUM has been around for 30 years, but what happened?" he quizzed during a presentation session at the IDF conference. It has been difficult to make OUM, because of the "difficulties with the materials," he answered.
So why is Intel paying so much attention to OUM? Intel--the world's leader in flash memory--believes that existing flash processes may run out of gas at the 45-nm (0.045-micron) technology node unless there are some fundamental changes implement within the cell structures in existing devices, according to Lai.
As a result, the Santa Clara-based chip giant is looking for a potential successor for flash and it claims OUM holds more promise over competitive technologies, such as FeRAM (ferroelectric random-access memory) and MRAM (magnetic RAM), Lai said. "We are looking at MRAM andFeRAM, but we think OUM has a better chance to succeed, as compared to the other technologies," he later told SBN.
In fact, Intel's interest in OUM dates back over three decades ago, when company co-founder Gordon Moore and the inventors of the memory technology, Energy Conversion Devices Inc. (ECD), co-authored a technical article on the subject in an issue of Electronics, a former publication at McGraw Hill Inc. The article appeared in the Sept. 28, 1970.
But 30 years later--in 2000--Intel raised eyebrows when the company took a stake in Santa Clara-based Ovonyx Inc. Intel is still working with the design house on commercializing OUM. Besides Intel, Europe's STMicroelectronics has also invested in Ovonyx.
Ovonyx is part of Troy, Mich.-based ECD, which in the late-1960s claimed it invented the world's first nonvolatile memory--called "Ovonic Electrically Erasable Programmable Read Only Memory."
But ECD failed to bring the technology into the commercial market, due to some technical problems, leaving some to wonder if OUM is more hype than substance, according to observers. But in 1999, ECD formed Ovonyx to help it achieve its 30-year goal of moving OUM technology to the marketplace.
Chips based on OUM store bits by generating different levels of low and high resistance on a glassy material. An OUM nonvolatile memory cell uses a horizontal strip of chalcogenide -- a type of electrically conductive glass material -- connected to an electrode.
When a high current is applied through the electrode, the chalcogenide heats to more than 600C in less than 10 nanoseconds, creating a region of amorphous glass with high resistance. When a lower current is applied, it heats up to less than 600C and then cools to a crystalline state with lower resistance.
OUM is also said to have several advantages over other futuristic nonvolatile memory types, but there's one problem: it is difficult to manufacturer and develop the calcogenide material, Lai said.
Still, the Intel executive believes there is some light at the end of the tunnel for OUM. Japan's Sony, Panasonic and others have made several breakthroughs in terms of material sciences, including the calcogenide material, he said.
The material is used in the "phase-change" mechanism in recordable CD-ROM devices, thereby solving at least one of the problems with OUM, according to analysts. "Material sciences have matured quite a lot," said analyst Jim Cantore of iSuppli Corp. in San Jose.
But, there are still other issues with OUM, including heat dissipation. "We have gotten one cell in OUM to work," he said. "The problem is to get one million cells to work together and put a billion devices into production," Lai explained.
So when will OUM finally live up to its promises and move into production? The Intel executive dropped hints that it could take three or more years to finish the work. He also noted that new and "disruptive" technologies do not appear overnight. "It took us about five years for us to develop flash from the concept stage to production," Lai said.
While the Intel executive seemed confident about the progress of OUM, Lai dropped hints that the technology may never move from drawing board to the production lines.
"We have a long ways to go and it could go anyway," he said. "But we have to take risks."